Design: Cluster-randomised, blinded, sham-controlled trial (68 operating rooms, 4-week blocks)
Population: 107,334 analysed adult operations lasting an hour or more; 58,938 inpatients
Primary outcome: 30-day composite of wound infection, infection-related complications or death (inpatients)
Effect: 8.1% vs 8.2%; OR 0.98 (95% CI 0.92–1.05); stopped for futility

ORACLE randomised 68 operating rooms in four-week blocks to active or identical-looking sham supplemental air purification units, with patients, clinicians and investigators blinded. The primary outcome was a 30-day composite of wound infection, infection-related complications or death after inpatient surgery lasting at least an hour.

The trial enrolled 110,538 operations and stopped at the third interim analysis for futility. In 58,938 inpatients the composite occurred in 8.1% with active purification and 8.2% with sham. Across all 107,334 analysed inpatients and outpatients it was 5.5% against 5.6%. The effect did not vary over time.

The trial was funded by the device manufacturer, which approved the original protocol but had no role in analysis, and it still found no benefit. The interval is narrow enough to exclude any meaningful effect.

For hospitals considering add-on air purification for theatres that already meet standard ventilation requirements, this is strong evidence to spend the money elsewhere. The authors suggest infection prevention is better aimed at direct wound contamination: skin preparation, glove changes, normothermia and timely antibiotic prophylaxis.

Do not buy supplemental theatre air purifiers to reduce surgical site infection.
Keep standard theatre ventilation and door discipline in place.
Focus infection prevention on skin preparation, antibiotic timing and normothermia.
Use this trial when reviewing infection-control purchasing proposals.

Why it matters

It removes a costly intervention from the infection-control list with unusually strong evidence.

The statistics, in plain English

An odds ratio of 0.98 with an interval of 0.92 to 1.05 means any effect, if one exists, is very small. Stopping for futility means the trial could not plausibly have shown benefit had it continued. Cluster randomisation by theatre and time block, with blinding, makes bias unlikely.